Google I/O 2026: AI search, AI agents, AI glasses, AI models

Jan Johannsen

20/5/2026

Translation: machine translated

At Google I/O, the focus is on software. Accordingly, there were many new AI announcements at the start, but also some hardware for AI.

The keynote speech at the start of Google I/O lasted two hours. There were plenty of new AI tools and models, combined with the realisation that using them costs money and requires a lot of trust in the supplier. Finally, there was also new hardware in the form of smart glasses.

Intelligent glasses will control AI from autumn

Google has high hopes for the glasses known as «intelligent eyewear». Google is supplying the software with Android XR and Samsung the technology. The first two models of glasses come from Gentle Monster and Warby Parker.

With these glasses, the AI can be used on the connected smartphone without having to pick it up. Built-in microphones are used for voice control and the cameras in the frame of the glasses allow the AI to see the surroundings and incorporate them or take photos.

Navigation, translations and everything else will be delivered to the ear via small speakers in the first models from the autumn. Later, there will also be models that can show information on a display in the glass. Further information on the price and availability of all models will follow in the coming months.

Fitbit Air was announced before I/O, but made many cameo appearances on wrists on stage.

New design for Gemini app

Of the many announcements, the innovations to the Gemini app are likely to be the most relevant for most people. «Neural Expressive» is what Google calls the new look, which includes new colours, haptic feedback and animations. The AI assistant also no longer provides text-only answers, but incorporates elements such as images, graphics, videos and timelines.

In addition, dialects will soon be selectable in the app. However, there were no German-language examples.

KI is not free

In view of the huge sums being invested in AI data centres, the use of AI cannot remain free of charge in the long term. For Google, this moment seems to have arrived. Although there are still numerous free basic functions, you have to pay for many of the new announcements.

For example, «Gemini Omni» can only be used by people who pay for AI Plus, AI Pro or AI Ultra. With the new model, new content can be created from text, videos, images and audio. It starts with videos in which the style, characters, environment or details can be edited using voice commands.

Also only available to paying users - in the USA - is the new «Daily Brief» in the Gemini app. The AI agent checks your inbox, calendar and other things every morning to tell you what's important - similar to «Now Brief» in Samsung's Galaxy AI.

If you want to create Google Docs by voice input with «Docs Live» or create or edit images in the workspace with «Google Pics», you also have to pay and the cheapest plan is no longer sufficient. AI Pro or AI Ultra are required.

For frequent users, AI is at least a little cheaper. Google is lowering the price of the Ultra plan - in the USA - from 250 to 200 dollars and is also introducing a 100-dollar version.

AI content should be recognisable

With SynthID, Google has a watermark for AI content that has labelled over 100 billion generated images and videos since its launch. Until now, however, content could only be checked via the Gemini app. In future, you should also be able to do this in Chrome, via Circle-to-Search or search.

With the introduction of the «C2PA Content Credentials Verification», the SynthID also contains more metadata. In addition, OpenAI, Kakao and IIIElevenLabs are other AI suppliers that support watermarking. Nvidia has been involved since last year.

The AI is looking for you

AI search is increasingly becoming the standard for Google search. With Gemini 3.5, it now uses a new AI model. Visually, this is visible through a new input screen that attempts to complete questions as they are entered

Later this summer, Gemini 3.5 Flash will provide dynamic layouts in the search results. The answers will then not only be in text form, but will also contain specially created graphics or animations if appropriate.

In order to be able to use AI agents in search, a paid plan is required. With Gemini Spark, you can then define parameters for stock prices, for example, and receive a notification when they are met.

With «Ask Youtube», even complex questions can be asked in the USA from the summer and the platform will find the right video for it. Similar to what is already possible with «Ask Maps» on Google Maps.

You have to trust the AI and Google

With Gemini Spark, Google is introducing an AI agent that is supposed to be your personal assistant. Like the hype tool OpenClaw, Spark should be able to do all kinds of things for you. Initially in the Gemini app, later also via email and in Chrome as well as Android Halo in Android 17.

In order to use Spark, you not only need an AI Ultra subscription from Google, you also need to give the AI agent access to your data. This starts with emails and calendar entries. Spark cannot complete its tasks without this data. For data that is already stored by Google, the hurdle should be lower than for external data. But OpenClaw has already shown that you shouldn't blindly trust an AI. After all, Google has built in a feature that allows Spark to confirm certain actions and not act without being asked.

With the Universal Commerce Protocol (UCP), Google is trying to establish a standard that makes it easier for AI agents to shop for you. In addition to new partners who support UCP in their shops, Google is expanding it to Canada, Australia and the United Kingdom in the coming months. In addition, not only products, but also hotels or food that can be ordered will be included in the UCP. With the Universal Cart, Google also wants to manage all of your orders.

Googlebooks and Android 17 with their own event

Since the two hours of the I/O keynote were reserved for AI, Google had already announced the new features of Android 17 and Googlebooks as the successor to Chromebooks a week earlier. It's not as if AI only played a minor role there.

Jan Johannsen

Senior Editor

Jan.Johannsen@galaxus.de

As a primary school pupil, I used to sit in a friend's living room with many of my classmates to play the Super NES. Now I get my hands on the latest technology and test it for you. In recent years at Curved, Computer Bild and Netzwelt, now at Digitec and Galaxus. more

---

News + Trends

From the latest iPhone to the return of 80s fashion. The editorial team will help you make sense of it all.

Show all

These articles might also interest you